FRACTAL ANALYSIS FOR CLASSIFICATION OF REGIONS IN OVERLAPPED FINGERPRINTS

G.B. VIDYADEVI¹*, H. SAROJADEVI², H.C. NAGARAJ^3
1Dept. of Computer Science and Engineering, Nitte Meenakshi Institute of Technology, Bangaluru- 560 064, Karnataka, India.
²Dept. of Computer Science and Engineering, Nitte Mahalinga Adyanthaya Memorial Institute of Technology, Nitte - 574 110, Karnataka, India.
³Dept. of Electronics and Communication Engineering, Nitte Meenakshi Institute of Technology, Bangaluru - 560 064, Karnataka, India.
* Corresponding Author : vgb2011@gmail.com

Received : 17-03-2015     Accepted : 30-04-2015     Published : 05-05-2015
Volume : 5     Issue : 1       Pages : 149 - 152
J Signal Image Process 5.1 (2015):149-152

Keywords : fractal, overlapped fingerprint, texture, naïve bayes classifier
Academic Editor : Nicholas Farmer, Mohammad Moradi
Conflict of Interest : None declared
Acknowledgements/Funding : Authors wish to thank the management, Nitte Meenakshi Institute of Technology, for encouraging the research work and providing kind support.

Segmentation of overlapped fingerprint is yet manually carried out by fingerprint examiners, which is a bottleneck in automation of overlapped fingerprints separation. This paper presents method for classification of regions in overlapped fingerprints into overlapped and non overlapped regions based on image fractal analysis. Overlapped fingerprints are decomposed into binary images using a combination multi Otsu thresholding and two threshold binary decomposition algorithms. Fractal dimensions are computed from border images derived from binary images using box counting method. The feature vector includes fractal dimensions, size of the object regions and their average gray value. Naive Bayes classifier is adopted for classification of overlapped fingerprints regions into overlapped and non overlapped regions. The results are evaluated on three databases, i) Standard simulated overlapped fingerprints database, ii) Real overlapped fingerprints database, and iii) Locally simulated overlapped fingerprints. The classification accuracy achieved is 88.33%, the misclassification is mainly due to poor quality of fingerprints in the non overlapped region.